Traffic detection system using machine vision

2. The traffic detection system of claim 1, wherein the traffic analytics engine further determines a lane detection region (LDR) mask for each lane based on analyzing some video frames received from the camera overlooking the roadway and based on an assumed average length of a vehicle; wherein the traffic analytics engine further processes the plurality of video frames, for each lane, with the LDR mask to produce the cropped video frames, wherein pixel intensity values of pixels of the cropped frames disposed outside the lane detection region mask are ‘0’ values.

3. The traffic detection system of claim 2, wherein each of the cropped video frames comprises a number of pixel intensity values that is less than twenty percent of the number of pixel intensity values in the video frames received by the traffic analytics engine.

4. The traffic detection system of claim 1, wherein traffic analytics engine determines the cropped background reference frame for each lane by averaging corresponding pixel intensity values of each of the background video frames.

5. The traffic detection system of claim 1, wherein the traffic analytics engine, for each lane, determines a mean pixel intensity value for each difference image associated with the lane; wherein the traffic analytics engine, for each lane, determines a pixel intensity variance for each difference image associated with the lane as the average of the squares of the differences found by subtracting the mean pixel intensity value determined for the difference image from each pixel intensity value of the difference image; wherein the traffic analytics engine, for each lane, determines a median pixel intensity variance from the pixel intensity variances determined from each of the difference images associated with the lane; wherein the traffic analytics engine, for each lane, determines the vehicle detection threshold for the lane by multiplying the median pixel intensity variance for the lane by a factor in the range of 2 to 50.

6. The traffic detection system of claim 1, wherein the traffic analytics engine, for each lane, determines a mean pixel intensity value for each difference image associated with the lane; wherein the traffic analytics engine, for each lane, determines a pixel intensity variance for each difference image associated with the lane as the average of the squares of the differences found by subtracting the mean pixel intensity value determined for the difference image from each pixel intensity value of the difference image; wherein the traffic analytics engine, for each lane, determines a median pixel intensity variance from the pixel intensity variances determined from each of the difference images associated with the lane; wherein the traffic analytics engine, for each lane, determines the vehicle detection threshold for the lane by multiplying the median pixel intensity variance for the lane by a factor in the range of 2 to 50; wherein the traffic analytics engine, for each lane, determines a windowed average pixel intensity value for each difference image for the lane as the average of the sum of the pixel intensity value for the difference image and a plurality of pixel intensity values for adjacent difference images for the lane; and wherein the traffic analytics engine, for each lane, identifies difference images for the lane that have the windowed average pixel intensity variance values that exceed the vehicle detection threshold.

7. The traffic detection system of claim 1, wherein the traffic analytics engine converts the video frames from color frames to black and white frames before generating the cropped video frames.

10. The method of claim 9, wherein each of the cropped video frames comprises a number of pixel intensity values that is less than twenty percent of the number of pixel intensity values in the video frames received by the traffic analytics engine.

11. The method of claim 8, wherein determining the background reference frame for each lane comprises averaging corresponding pixel intensity values of each of the background video frames.

13. The method of claim 8, further comprising converting the video frames from color frames to black and white frames before generating the cropped video frames.

14. The method of claim 8, further comprising thinning out pixel intensity values of the video frames by the traffic analytics engine before analyzing generating the cropped video frames.

15. The traffic detection system of claim 1, wherein the cropped video frames used to determine the background frames are successive cropped video frames.

16. The method of claim 8, wherein the cropped video frames used to determine the background frames are successive cropped video frames.